For several decades, the National Institutes of Health has recognized the importance of developing a diverse workforce of biomedical researchers. Yet relatively little is known about the reasons that attract URM medical students and physicians to enter research careers. The proposed study has the primary aim of addressing this knowledge gap, with the broader goal of informing interventions to address this national need. The project will bring together a diverse, interdisciplinary team of researchers and educators with great depth and breadth of experience. In contrast to prior non-theoretical research efforts, we propose to apply a well-respected social science model, the Theory of Planned Behavior, to guide this research. The project has two parts: 1) a prospective study of four cohorts of Harvard Medical School (HMS) students from the entering classes of 2009-2012 (n=660), and 2) a retrospective study of four cohorts of HMS graduates after they have completed all their training, the graduating classes of 1996-1999 (n=703). The prospective study takes advantage a previously planned curricular change at HMS: With the entering class of 2011, all HMS students will be required to complete a mentored research project. We propose to study the last two cohorts entering under the old requirement and the first two cohorts under the new. Moreover, we will survey each of the four cohorts longitudinally, at the end of their pre-clinical basic science education and at the end of their principal clinical year. The primary outcome will be percentage of effort devoted to research, in the form of intentions for students and reported actual effort for graduates. We predict that identification with the HMS culture will be closely related to research career choice for URM students; that research career intentions will be higher for all students under the new requirements; and that the factors that affect students will be different at the end of their second year versus third year. The project has sufficient statistical power to address the major aims. To overcome an inherent limitation of the prospective study (intentions rather than actual choices can be studied), the retrospective survey will be conducted in parallel on graduates of the same medical school. By having multiple data sources (students and graduates) about motives for choice, as well as data about both intentions and actual choices, we can cross-validate our findings and offer a clearer picture of the determinants of research career paths as affected by racial/ethnic background, attitudes, and prior experience. Our findings will guide future programs on how to attract more URM medical students into research careers in biomedical and behavioral sciences.